NBER WORKING PAPER SERIES PRIVATE CREDIT IN 129 COUNTRIES. Simeon Djankov Caralee McLiesh Andrei Shleifer

Transcription

1 NBER WORKING PAPER SERIES PRIVATE CREDIT IN 129 COUNTRIES Simeon Djankov Caralee McLiesh Andrei Shleifer Working Paper NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA January 2005 The views expressed herein are those of the author(s) and do not necessarily reflect the views of the National Bureau of Economic Research by Simeon Djankov, Caralee McLiesh, and Andrei Shleifer. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including notice, is given to the source.

2 Private Credit in 129 Countries Simeon Djankov, Caralee McLiesh, and Andrei Shleifer NBER Working Paper No January 2005, Revised June 2006 JEL No. G3, G32, K22 ABSTRACT We investigate cross-country determinants of private credit, using new data on legal creditor rights and private and public credit registries in 129 countries. We find that both creditor protection through the legal system and information sharing institutions are associated with higher ratios of private credit to GDP, but that the former is relatively more important in the richer countries. An analysis of legal reforms also shows that improvements in creditor rights and in information sharing precede faster credit growth. We also find that creditor rights are extremely stable over time, contrary to the convergence hypothesis. Finally, we find that legal origins are an important determinant of both creditor rights and information sharing institutions. Simeon Djankov World Bank Research Department 1818 H Street, NW Washington, DC sdjankov@worldbank.org Caralee McLiesh World Bank Group International Finance Corporation 2121 Pennsylvania Avenue, NW Washington, DC cmcliesh@worldbank.org Andrei Shleifer Department of Economics Harvard University M9 Littauer Center Cambridge, MA and NBER ashleifer@harvard.edu

3 I. Introduction Recent research in the development of private credit markets across countries points to an important role of institutions, such as legal investor protection and credit bureaus, in supporting these markets (e.g., La Porta et al. 1997, Japelli and Pagano 2002). Although this research has identified several empirical regularities, it suffers from small samples, weak results, and some econometric problems. This paper greatly expands the available data set on credit institutions to cover 129 countries and 25 years of data. The expansion allows an advance on previous research in five directions. First, the greater country coverage yields statistically more reliable results on the role of alternative credit institutions in a cross-section. Second, the larger sample allows us to examine whether different institutions are more effective in countries at different levels of economic development and judicial efficiency. Third, time series data on institutions allow us to ask whether these institutions improve over time and/or converge across countries. Fourth, the new data allow us to analyze the consequences of institutional reforms for financial development, thereby avoiding some econometric concerns about cross-sectional work. Fifth, the much larger sample allows us to better distinguish the roles of legal origins and culture in shaping credit institutions (Stulz and Williamson 2003). There are two broad views of what determines how much private credit a financial system would extend to firms and individuals. According to the first, what matters for the viability of private credit is the power of creditors. When lenders can more easily force repayment, grab collateral, or even gain control of the firm, they are more willing to extend credit. These power theories of credit have been formalized by Townsend (1979), Aghion and Bolton (1992), and Hart and Moore (1994, 1998). According to the second view, what matters for lending is information. When lenders know more about borrowers, their credit history, or other lenders to the firm, they are not as concerned about the lemons problem of financing 2

4 non-viable projects, and therefore extend more credit. These information theories of credit have been pioneered by Jaffe and Russell (1976) and Stiglitz and Weiss (1981). We study the importance of information and power theories of credit in explaining the variation in the size of private credit markets around the world using a new data set on private credit institutions for 129 countries during the period To assess the power theories of credit, we construct a measure of legal rights of creditors in these countries, the creditor rights index first proposed by La Porta et al. (1997, 1998), for every year during this period. The index measures the legal rights of creditors against defaulting debtors in different jurisdictions, and has been previously interpreted as a measure of creditor power. To assess the information theories of credit, we collect data on the existence of public (i.e., governmentowned) and private credit registries in different countries during the same period. These registries collect information on credit histories and current indebtedness of various borrowers, and share it with lenders. Credit registries exist in many countries, and have been shown to be an important factor in determining credit availability (Jappelli and Pagano 2000, 2002, Pagano and Jappelli 1993, Sapienza 2002). Our initial empirical strategy is to run cross-country regressions that explain the private credit to GDP ratio in terms of creditor rights and the presence of registries. In addition, we take advantage of the time series dimension of the data. First, we ask how stable the institutions of credit are over time and whether they converge or diverge among different groups of countries. In particular, we test the hypothesis that countries whose laws derive from different legal traditions exhibit convergence in their creditor rights scores. Second, by looking at the changes in either creditor rights or the information institutions, we can ask whether these reforms have any effect on private credit. This differences-in-differences approach gets around the concern that credit institutions are endogenous, and presents an alternative to instrumental variable techniques that have also been criticized (e.g., by Glaeser et al. 2004). 3

5 Creditor power and information theories are not mutually exclusive. Both ex ante (and interim) better information and ex post stronger creditor rights can contribute to credit market development. Indeed, as suggested by Jappelli and Pagano (2002), these institutions may be substitutes: some countries may specialize in information institutions, others in legal systems giving power to the creditors. We explore the substitution hypothesis in some detail. First, less developed countries, with poorly functioning legal systems, might be unable to sustain an effective lending channel based on ex post creditor rights, and may depend on information sharing for their credit markets to function. In contrast, richer countries might develop more functional systems of bankruptcy, so that creditor power can be particularly important in these countries. We examine the relative importance of information and power theories for countries at different levels of development. Second, as argued by Djankov et al. (2003a) and Mulligan and Shleifer (2005), countries in different legal traditions may specialize in different strategies of social control of business. Specifically, common law countries often rely on private contracting, whereas civil law countries, and especially French civil law countries, rely on government regulation and ownership. Credit markets present an opportunity to evaluate these predictions, since creditor power theories deal with legal creditor rights and their enforcement through courts, whereas information theories deal with mandatory information disclosure and often state-owned institutions for information sharing. We further distinguish between private and public credit registries, and examine the roles of both creditor rights and these information sharing institutions in supporting credit markets in countries from different legal traditions. Our results can be briefly summarized. First, we find a pronounced legal origin effect in credit market institutions, with common law countries having sharply higher creditor rights scores than French civil law countries. The latter, in contrast, have a much higher incidence of public credit registries than do the former. Second, we find very little convergence in creditor 4

6 rights scores, or in information institutions, among legal origins. At least for these measures of legal institutions, the differences persist over the 25 year period. Third, we find that both the creditor rights scores, and the incidence of private credit registries, are higher in the richer than in the poorer countries. Fourth, we find in a cross-section of countries that both better creditor rights and the presence of credit registries are associated with a higher ratio of private credit to GDP. In addition, we find that private credit to GDP ratio rises following either improvements in creditor rights or the introduction of credit registries. The confirmation of cross-sectional evidence with an analysis of reforms is a significant innovation in this paper. Fifth, we find that creditor rights are particularly important for private credit in the richer countries, whereas credit registries matter in the poorer countries. These results both strengthen and refine earlier findings, but also point to a beneficial role of public credit registries in poor French legal origin countries a rare example of an apparently successful state intervention. The next section presents our data. Section III presents the basic results on the effects of various institutions on private credit. Section IV looks at the effects of changes in creditor rights, and of the introduction of private and public registries, on private credit. Section V examines the variation in the prevalence of these institutions across countries. Section VI concludes. II. The Data A. Variable Definitions We gather data on 133 countries, comprising every economy with a population over 1.5 million, except countries in civil conflict or inactive members of the World Bank, such as Iraq, Afghanistan, Myanmar, Cuba and Sudan. Table 1 describes the variables used and their sources. The private credit data is from the IMF s International Financial Statistics, lines 22d and 42d, which measure claims on the private sector by commercial banks and other financial institutions. The variable is expressed as a percentage of GDP. Data are available for all sample 5

7 economies except Puerto Rico, Serbia and Montenegro, and Uzbekistan. We exclude China, since the credit variable includes credit to state-owned enterprises and stands at a staggering 130% of GDP. In comparison, the share of private credit to GDP in the United Kingdom is 136%, in Germany 118%, and in France 87%. This leaves us with a maximum of 129 countries in the analysis of private credit. The creditor rights index follows that constructed by La Porta et al. (1997), with minor differences. We construct the index as at January for every year between 1978 and 2003, and expand their sample from 49 to 133 countries. The creditor rights index measures four powers of secured lenders in bankruptcy. First, whether there are restrictions, such as creditor consent, when a debtor files for reorganization. Second, whether secured creditors are able to seize their collateral after the petition for reorganization is approved, in other words whether there is no automatic stay or asset freeze imposed by the court. Third, whether secured creditors are paid first out of the proceeds of liquidating a bankrupt firm. Finally, whether an administrator, and not management, is responsible for running the business during the reorganization. A value of one is added to the index when a country s laws and regulations provide each of these powers to secured lenders. The creditor rights index aggregates the scores and varies between 0 (poor creditor rights) and 4 (strong creditor rights). Our creditor rights data as of 1995 are the same as La Porta et al. (1997) report for over 90% of the observations. The most significant differences arise from coding different bankruptcy procedures. For India, for example, La Porta et al. code using the reorganization provisions in the Companies Act (1956), while we code with respect to the Sick Industrial Companies Act (1985), which is more relevant for industrial firms. Australia, Pakistan and Sri Lanka are similar. In some instances, such as Malaysia and the Netherlands, differences in coding arise because of the treatment of the power of the administrator in reorganization. We code that management does not stay if an administrator is automatically appointed and the debtor does NOT remain in 6

8 control in the ordinary course of business. In a few cases, differences derive from the moment at which there exist restrictions on entering reorganization. We code with respect to restrictions at the gate, i.e., the initial point of entering reorganization. In 2003, twenty one countries, including Benin, Chad, Colombia, France, and Tunisia, had a score of 0. Nine countries, including Hong Kong (China), Kenya, Lebanon, New Zealand, Panama, and the United Kingdom, had a perfect score. The United States has a score of 1, as secured creditors are paid first out of the proceeds of bankruptcy, but there are no restrictions on entering reorganization, the debtor benefits from the automatic stay on assets, and management runs the company during the reorganization process. We collect time series data on creditor rights in two stages. We begin with a review of bankruptcy and bankruptcy-related laws from 1978 to 2003, identifying all major reforms and assessing their impact on the creditor rights index. We then survey local bankruptcy lawyers, confirming the dates of reforms and their impact on the creditor rights index. We review bankruptcy law and reforms primarily using online legal resources, including Foreign Law Guide, LexisNexis database, online library of the International Bar Association, European Restructuring and Insolvency Guide, and Asia-Pacific Restructuring and Insolvency Guide. 1 Additional legal resources include Collier International Business Insolvency Guide and International Insolvency. 2 In identifying the number of reforms of bankruptcy law, we exclude changes that affect bankruptcy only for financial institutions, state enterprises, or persons. Likewise, purely procedural law reforms are not considered. We record all reforms to bankruptcy legislation i.e., reorganization, liquidation, foreclosure, and secured transactions laws. Substantive bankruptcy regulations, however, are not 1 Foreign Law Database: International Bar Association: The Asia- Pacific Restructuring and Insolvency Guide 2003/2004: The European Restructuring and Insolvency Guide 2002/20003: 2 Collier International Business Insolvency Guide published by Matthew Bendar and Company Inc. a member of the Lexis- Nexis Group. International Insolvency published by Juris Publishing (2002). 7

9 always contained in the Bankruptcy Law. Thus, we keep track of reforms of Corporate Law, Company Law, Commercial Law, the Civil Code, and other associated laws if such reforms affect bankruptcy. As an example, take Malawi. In 2000, the passage of the Employment Act, a law seemingly unrelated to bankruptcy matters, affected the ranking of secured creditors claims in liquidation. Before the reform, the Companies Act of 1984 ranked secured creditors first. The 2000 Employment Act ranked claims for wages ahead of secured creditor claims. The reform lowered Malawi s creditor rights index from 3 to 2. We survey a total of 440 lawyers from the 133 countries to verify the results of the legal review. We ask our respondents to either confirm or amend our initial findings on the number and timing of reforms affecting bankruptcy since In addition, we verify whether and how the reforms impacted the creditor rights variables. In six countries the respondents amended our findings, generally to correct the timing of the reform to reflect when it came into force. Since 1978, many countries reformed their bankruptcy laws. By our count, 99 countries had at least one reform, and the total number of reforms is 162. We found evidence that richer countries have more reforms, inconsistent with the idea that poorer countries reform more to catch up with the richer ones. However, few reforms affected the La Porta et al. creditor rights index. Only 25 countries had changes in their creditor rights index since 1978, with a total of 32 changes. The correlation between the 2003 and 1978 creditor rights indices is Appendix A lists all bankruptcy law reforms in this sample. Appendix B reports all changes in creditor rights scores. We record the presence of public and private credit registries though a survey of banking supervisors. Public credit registries are databases managed by a government agency, usually the Central Bank or the Superintendent of Banks, that collect information on the standing of borrowers in the financial system and make it available to actual and potential lenders. In 2003, they operated in 71 countries in our sample. Some, such as the German and Saudi Arabian 8

10 registries, collect only limited information on outstanding loans of large borrowers, and focus on banking supervision. Others, such as those in Belgium, Ecuador, Malaysia, and Taiwan distribute extensive information including on late payments and defaults, demographic data, credit inquiries, ratings, and sometimes even the payment of utility bills and court records of the company and its owners. A private credit bureau is a private firm or non-profit organization that maintains a database on the standing of borrowers in the financial system. Its primary role is to facilitate exchange of information among banks and financial institutions. 3 As of 2003, private bureaus operated in 55 of our sample countries, including all OECD countries but France. Three international firms Experian, Equifax and TransUnion either own or are affiliated with half of the bureaus in our sample. Unlike public registries, private bureaus usually gather information from non-bank lenders and public sources, distribute more data, and offer a broader range of services to lenders. The New Zealand bureau, for example, offers credit scoring, borrower monitoring, fraud detection, debt collection and marketing services. For countries that confirmed the presence of a public or private registry, we conducted a detailed survey of its structure, laws, and associated rules. The survey was filled by the director of the registry or, in cases where the registry is a department of the Central Bank, by the department s head. To build the time series data, respondents were asked the year in which the registry was legally established, as well as the year in which the registry began operations i.e., distributing credit information. In some cases, the difference is significant. For example, the Turkish credit bureau was incorporated in 1995, but did not become operational until 1999, following years of negotiating with data providers and developing technology. The analysis is based on the year in which the registry began operations. 3 Credit investigative bureaus and credit reporting firms that do not directly facilitate exchange of information between financial institutions exist in many countries, but are not considered here. 9

11 In addition to laws on the books, the quality of law enforcement is likely to matter. We control for enforcement with a measure of the number of days it takes to enforce a simple debt contract. Data are based on the methodology developed in Djankov et al. (2003b), with one change, namely that the current series refer to the time to enforce a contract of unpaid debt worth 50% of the country s GDP per capita as of January This amounts to about $18,000 in the United States, about $12,000 in France and Germany, $6,000 in Korea, and $3,000 in Mexico. We expanded the Djankov et al. data set to 133 countries. 4 It takes only 48 days to enforce a debt contract in the Netherlands, 50 days in New Zealand, 69 in Singapore, 60 in Japan, and 75 in Korea. In contrast, it takes 1,459 days to enforce a debt contract in Guatemala, 1,028 in Serbia, 1,003 in Slovenia, and 1,390 in Italy. We control for each country s total GDP, as it has been suggested that larger economies may have bigger credit markets because of economies of scale in organizing the supporting institutions. We also control for growth of GDP, because rapid economic expansion may require more credit. The GDP data come from the World Bank s World Development Indicators (CD- Rom, January 2004), and like private credit data are not available for many countries in the earlier part of our period 5. A country s legal origin has been shown to be an important determinant of both creditor rights and private credit (La Porta et al. 1997, 1998; Levine 1999; Beck et al 2003a, b). There are four main legal origins: English, French, German, and Nordic. The English legal origin includes the common law of England, and the colonies to which it spread, including the U.S., Australia, and Canada. The French legal origin includes the civil law of France, of countries Napoleon conquered (including Portugal and Spain), and of their former colonies. The German 4 The correlation between our variable and the Djankov et al. (2003a) measure for the overlapping sample of 88 countries is Data on the ratio of private credit to GDP are not available for 497 out of 3225 potential observations, mostly the transition economies during the communist period. In these cases, we only use country-years with available private credit to GDP ratios in the regressions. 10

12 legal origin includes the laws of the Germanic countries in Central Europe, but also countries in East Asia where the German law was transplanted. The Nordic legal origin refers to the laws of the four Scandinavian countries. We use this La Porta et al. classification and add a fifth category: Socialist (transition). The countries in this category have inherited Soviet laws: these include 12 countries that emerged from the breakup of the Soviet Union, plus Mongolia. We do not apply the Socialist category to countries that have gone back to their pre-communist legal systems. Specifically, the Baltic and East European countries reactivated their pre World War II laws after the collapse of communism (these laws always stayed on the books, but were not used during the communist era), and so we classify them according to the origin of those laws. Latvia s laws belonged to the German civil law tradition prior to annexation by the Soviet Union in 1940; it reverted back to them in Lithuania was influenced by French and Dutch law both before its annexation in 1940 and after independence in It is classified as French legal origin. The remaining former communist countries in central and Eastern Europe Bulgaria, Croatia, the Czech Republic, Hungary, Poland, Serbia, the Slovak Republic, and Slovenia - followed the German legal tradition, with the exception of Romania, which followed the French tradition, and Albania, which inherited French legal influences via Italy. These countries as well are assigned to their pre-war legal systems 6. Finally, it has been argued that religion is an important determinant of credit institutions (Stulz and Williamson, 2003). We record the religion practiced by the largest proportion of the population, as recorded in the CIA factbook (December 2003, online edition). 6 The former communist countries only began reporting private credit to the IMF during , so the question of applicable legal origin during the communist era does not arise. We should also emphasize that this allocation of countries to legal origins is based on their bankruptcy laws, and might be different for other business laws. 11

13 B. Summary of the Data Table 2 presents the data on credit institutions around the world by legal origin. Panel A shows creditor rights, Panel B public registries, Panel C private bureaus, and Panel D the combined information sharing variable, defined as a dummy equal to 1 if a country has either a public or a private registry. We present the data at five year intervals between 1978 and Panel A confirms for a much larger sample of countries the finding of La Porta et al. (1997, 1998) that investor protection through creditor rights varies systematically across legal origins, but in particular is higher in common law than in French civil law countries. This result is highly statistically significant, and is as strong in 1978 as in The Table also confirms the earlier finding that German civil law countries have strong creditor rights and Nordic countries weak ones, although not as weak as those of French legal origin countries. In addition, Panel A demonstrates that there has been no trend toward greater creditor rights over time in any legal origin, with the possible exception of the socialist countries. Indeed, there is almost no change in average creditor rights score over time in any legal origin. This implies that there has been NO CONVERGENCE of creditor rights between legal origins: the French and common law countries are as far apart in 2003 as in The stability of creditor rights scores over time, and the absence of convergence across legal origins, is broadly consistent with the view that these particular measures of investor protection reflect relatively permanent features of the institutional environment, deeply rooted in national legal traditions. Panel B presents the results on public registries. It shows that these registries are more common in French and German civil law countries than in Nordic and common law countries. In 2003, public registries existed in 76.6 percent of French legal origin countries, 61.1 percent of German legal origin countries, a quarter of common law countries, and no Nordic countries. The difference between French and common law countries is statistically significant in 2003, but also in every prior year in the Panel. 12

14 Unlike the results of Panel A, which show no time trend, Panel B shows substantial increases in the incidence of public registries in all legal origins but the Nordic. In 1978, there were almost no public registries except in French legal origin countries; by 2003, over half the countries in the world had them. The results of Panels A and B suggest that public registries may substitute for creditor rights, at least comparing common law and French civil law countries. Panel C reports the results for private credit registries. They are somewhat more frequent in common law countries than in French legal origin countries, and by 2003 universal in Nordic countries, but the difference between French and English origins is not statistically significant. Like the public registries, private bureaus have become much more common around the world in the last 25 years, and the ranking of legal origins is generally preserved. Panel D shows that some kind of an information sharing institution (i.e., public or private) occurs more frequently in civil than in common law countries. As we saw from Panels B and C, however, this is driven by public registries rather than private bureaus. The bottom line is that there are pronounced legal origin effects in creditor rights and in public registries, going in opposite directions. However, creditor rights are extremely stable over time, whereas both public and private information sharing institutions are becoming increasingly common during the time period Table 3 looks at the same data arranging countries by income level in Many countries do not have data for GDP per capita in 1978, but the results we report are very similar if we use 2003 (or an interim year) per capita income rankings. Panel A shows that, unlike in the La Porta et al. data, richer countries have statistically significantly higher creditor rights scores than do poorer countries, and that this is true at all points in time. Part of this of course is the composition effect, as Nordic and German legal origin countries are both rich and have high creditor rights, so we need to run regressions to validate this finding. Panel A also shows that in neither the richer nor the poorer countries has there been a change of creditor rights over time. 13

15 Panel B shows that the poorer countries have a higher incidence of public registries than do the richer countries; in contrast, Panel C shows that the richer countries have a much higher incidence of private bureaus. If we divide the countries by both legal origin and income, we find that, in 2003, 83.3 percent of the poor French legal origin countries have public credit registries, compared to only 26.3 percent of the poor common law countries a truly striking difference. The results of Tables 2 and 3 are suggestive of possible patterns of substitution among credit institutions. First, in the domain of information sharing, richer countries rely relatively more on private institutions, and poorer countries on public ones. Second, common law countries appear to emphasize the ex post mechanism of creditor power over the ex ante mechanism of information sharing, with the reverse holding for civil law countries. The conspicuous exceptions to this tradeoff are the German civil law countries, which have both strong creditor rights and extensive information sharing institutions. Third, looking at the information sharing institutions, French legal origin countries specialize in public registries relative to countries in other origins. This specialization is most conspicuous in poor countries. Table 4 presents correlations among our variables. First, it documents a positive correlation between the amount of private credit and creditor rights, the existence of private credit bureaus, aggregate income, income per capita, and income growth, as well as English, German and Nordic legal origins. Private credit is negatively correlated with French and socialist legal origin, as well as with the inefficiency of the legal system as measured by contract enforcement days. Second, Table 4 confirms the suggestion of Table 2 that creditor rights and public registries are substitutes, and that public and private registries might be substitutes as well. Public registries also appear to be more prevalent in countries with less efficient judicial systems. We revisit this evidence on institutional substitution in the next section. 14

16 III. Cross-Country Determinants of Private Credit In this section, we present cross-sectional results on the determinants of private credit to GDP ratio in 129 countries in We have replicated these results for the cross-sections every five years from 1978 to 1998, and the results we report hold in other time periods as well. Before presenting the results, three points should be noted. First, we try to present the results in the simplest and most transparent form. We have estimated other specifications, including various interaction terms, and the conclusions we draw are confirmed. Second, we discuss a number of robustness checks below, and we have tried others. However, most potential control variables are available for only a small subset of countries, so adding them often results in cutting the sample by two thirds. Third, although an argument can be made that both legal rules and information sharing institutions are endogenous, we present in Tables 5 and 6 ordinary least squares results. Following La Porta et al. (1997), we have confirmed these results using legal origins as instruments, but instrumental variables techniques have significant problems in this context, including the fact that both of the key independent variables vary by legal origin. In Section V below, we examine some determinants of private credit institutions, including legal origins (see Table 8). Table 5 examines the effect of creditor rights on private credit in a framework similar to La Porta et al. (1997), except using a much larger cross-section of countries (129 versus 39). The dependent variable is the average ratio of private credit to GDP, expressed as a percentage, computed over The creditor rights score is as of We use several controls in these OLS regressions. First, as in La Porta et al. (1997), we control for the total GDP on the theory that credit markets might require fixed institutional costs to function, which are only paid when the total economy is large enough. Second, also as in La Porta et al. (1997), we control for per capita GDP growth, since faster growing economies may have greater demand for credit. Third, we use our updated measure of days to enforce a simple contract as a proxy for the 15

17 efficiency of the legal system. Presumably, debt contracts are more prevalent in the more efficient legal systems. Fourth, since inflation may devalue the stock of outstanding debt or otherwise undermine debt contracting, we control for contemporaneous inflation. The first column shows that GDP, GDP per capita growth, and contract enforcement days (but not inflation) enter significantly and with theoretically predicted signs. In particular, the objective measure of the quality of courts is a significant predictor of private credit. Further, countries with stronger legal protection of creditors have deeper credit markets (Figure 1). As the creditor rights index rises by 1 (roughly the difference between the French and the German legal origins), the private credit to GDP ratio rises by 6 percentage points (or a fifth of the way from the French to the German legal origin mean). As do La Porta et al. (1997), we find support for the power theories of debt, in which creditor rights determine the willingness to extend credit. However, the results presented here are substantially stronger than those presented by La Porta et al. (1997) for a sample of 39 countries as of Part of the reason for this is that several countries changed their bankruptcy scores between 1995 and 2003 in a way that improved the fit. Yet even if we rerun the La Porta et al. (1997) regression using their 39 countries but our data as of 1995, we get stronger results than they do, indicating that recoding the variables for a few countries, as discussed in Section I, improves the fit. The second column in Table 5 includes legal origins as additional controls. Except for the fact that transition countries have less developed debt markets, there is no significant influence of legal origin on private credit beyond that contained in other variables. In fact, the coefficient on the creditor rights index barely changes. There is no additional information in legal origins for the effect of legal rules on private credit markets. The next four columns revisit these results for the richer and the poorer countries separately. Note first that the quality of courts measure is only significant for the richer countries, which suggests that the legal mechanism enforcing debt contracts is not important in 16

18 the poorer ones. Note second that inflation in these specifications exacts an adverse effect on debt markets, especially in the richer countries. Turning to our main results, the evidence makes clear that the creditor rights results are driven by the richer countries. For these countries, both contract enforcement days and creditor rights remain significant determinants of private credit (with legal origins remaining unimportant). For the poorer countries, in contrast, the statistical significance of these variables disappears. This result again suggests that the legal mechanism of enforcing debt contracts is more important in the richer countries. We have examined which components of the creditor rights index are responsible for its ability to predict private credit. We find that the absence of automatic stay on assets and respect for the priority of secured creditors matter a great deal. On the other hand, restrictions on entering reorganization and mandatory removal of management in bankruptcy are not particularly important. This evidence suggests that the power to grab and liquidate collateral by secured creditors supports successful debt markets. 7 We check the robustness of the results in Table 5 in several ways. First, following Mulligan and Shleifer (2005), we replace the overall GDP by population as the scale variable. This would make sense if both the benefits and the costs of running institutions rise with per capita income. All of the results we have just described are preserved when we make this substitution. Second, we replace our preferred measure of the efficiency of the judicial system, contract enforcement days, with the more conventional GDP per capita. We find, as in Table 5, that countries with higher per capita income have a higher ratio of private credit to GDP, and that this result is driven by the rich rather than the poor countries. The coefficients on other variables of interest do not change materially. Third, instead of dividing countries into rich and poor, we divide them according to their ICRG law and order score, on the theory that it is the law and 7 In a related vein, Qian and Strahan (2004) find that, in a cross-section of countries, a higher creditor rights index is associated with a higher likelihood that bank loans are secured. 17

19 order environment rather than just development that determines the relevance of specific legal rules. We find, as with per capita income, that creditor rights matter in the high but not in the low law and order countries. Fourth, it can be argued that our private credit variable includes too much credit to state enterprises. Accordingly, we replace it by the private bond market capitalization measure from Beck et al (2000), which unfortunately is only available for 36 countries. Even in this small sample, the creditor rights score is a statistically significant predictor of private bond market capitalization. Fifth, we control individually for a number of variables in the regression that might be correlated with the creditor rights score, but also influence the development of private credit. These include the ratio of stock market capitalization to GDP, a measure of the importance of state enterprises in the economy, a measure of fiscal deficit, several proxies for central bank independence, and several proxies of judicial checks and balances. 8 None of these variables change our results. Finally, we considered the possibility that not the creditor rights index itself, but some other aspect of legal or regulatory intervention in debt markets correlated with creditor rights, is responsible for credit market development. Specifically, we include several measures of the quality of bank supervision assembled by Barth, Caprio, and Levine (2004). While some of these variables are statistically significant, they do not eliminate the statistical significance of creditor rights. In Table 6, we add the three measures of information sharing the existence of a public registry, that of a private bureau, and that of either (information sharing = 1 if either a public registry or a private bureau exists in a country) to the regressions in Table 5. Like creditor rights, these variables are measured as of We eliminate legal origins from the specifications, since they do not influence private credit holding creditor rights constant. 9 The effects of the creditor rights index and of contract enforcement days do not change much at all 8 For motivation for inclusion several of these controls, see Beck, Demirguc-Kunt, and Levine 2003a, 2003b,

20 from the estimates in Table 5 in all six specifications. In addition, the data show that public registries are associated with more private credit, but only in the poorer countries (Figure 2), and that private credit bureaus are associated with more private credit in both the poorer ones and the overall sample (Figure 3). When we combine these public and private registries into the information sharing variable, it has a statistically significant and quantitatively large effect on private credit in both the poorer countries and the overall sample. As noted earlier, credit registries vary significantly in their design. We analyze the extent to which particular characteristics of registries are associated with more private credit with a detailed survey of the structure, rules, and governing laws of credit registries. We find that registries that distribute a broader range of data and provide legal incentives to ensure quality are associated with significantly more private credit. We identify six characteristics that in the data encourage private credit: 1) both positive information, meaning loans outstanding and payment history on accounts in good standing, and negative information, meaning defaults and arrears, is distributed; 2) data on both firms and individual borrowers is distributed; 3) data from retailers, trade creditors, and/or utilities, as well as from financial institutions, is distributed; 4) five or more years of historical data is available; 5) data are collected on all loans of value above 1 percent of income per capita; and 6) laws provide for borrowers right to inspect their own data. Many of these characteristics might be endogenous, and so we merely note these results, rather than interpret them as causal. The results in Tables 5 and 6 confirm some of the patterns of institutional substitution identified earlier. Whereas creditor rights and judicial efficiency influence private credit in the richer countries, neither public nor private credit registries do. In contrast, both public and private registries benefit private credit in poorer countries. The data thus suggest that the power 9 The results are robust to the inclusion of legal origins. Also, we performed the same robustness checks for the results in Table 6 as we did for those in Table 5, and the results remain robust. 19

21 mechanism for sustaining credit works in the richer countries, while the information mechanism matters in the poorer ones. 10 Moreover, there appears to be a constructive role for government in maintaining public credit registries in the poorer countries, where for reasons of cost or compliance private credit registries are uncommon (see Table 3). Table 2 suggests that governments in poor French legal origin countries are more likely to play this constructive role. These results identify a role for both the power and the information theories of debt, but under different circumstances. IV. An Analysis of Reforms Cross-country regressions are often criticized because they may omit important country characteristics or because the independent variables are endogenous. An alternative empirical strategy is to look at the change in the private credit to GDP ratio around institutional reforms that either change the creditor rights score or introduce a public registry or a private bureau. We do this in two alternative ways recommended by Bertrand, Duflo, and Mullainathan (2004), and report the results in Table 7. The first strategy, pursued in Panel A, is to collapse the data. Specifically, we restrict the sample to countries that undertook reforms, and run a cross-sectional regression of the change in the average levels of private credit to GDP ratio after versus before the reform on the change in institutions. For concreteness, begin with reforms of creditor rights. Suppose country i changes its creditor rights score in year t. Let AC (i, before) be the average private credit to GDP ratio of the reforming country i taken over the five years before year t, and AC (i, after) be the average private credit to GDP ratio in this country in the five years after the reform year t. Let AC (i) = AC (i, after) AC (i, before) be the change in the average credit to GDP ratio of the reforming 10 An important caveat to this interpretation is that the La Porta et al. index of creditor rights is computed from the perspective of secured creditors in bankruptcy, and that this particular measure of investor protection as opposed 20

22 country i following the reform. We are interested in how AC (i) is affected by the change in the creditor rights score in country i (occurring in year t), CR (i). However, we need to worry that, during the period t-5,,t,,t+5, the private credit to GDP ratio changed systematically in non-reforming countries as well. To this end, for each country i reforming in year t, consider all countries j = 1,, J which did not have a creditor rights change in the years t-5,,t,,t+5, i.e., the whole ten year window around the year t of reform in country i. For each of these countries, compute AC (j) as above, i.e. the change in private credit to GDP ratio in the 5 year window around the year t of reform of country i. Finally, compute AC (control, i) as the average AC (j) over all the countries j not reforming in the window around t. This AC (control, i) can be computed for every reforming country i, and will be used as a control variable. Obviously, AC (control, i) differs for every year. We run on the cross-section of only the countries that actually experienced a change in creditor rights score the following ordinary least squares regression: AC (i) = α + β* CR (i) + γ* AC (control, i). The coefficient γ captures world trends in private credit to GDP ratio around reform years, while the coefficient β captures the effect of a unit change in the creditor rights score on private credit. The same regression can be run for 3 year windows around creditor rights score reforms. For the public and private registries, the change in institutions the counterpart of CR (i) -- is always equal to one, since we only have the introductions of private bureaus or public registries in the data. Accordingly, we follow the procedures above (including the calculation of the corresponding control variables) for introductions of private and public registries (rather than changes in creditor rights score), and then run AC (i) = α + γ* AC (control, i), to creditor rights more generally, might be especially relevant to debt finance in the richer countries. 21

23 on the sub-samples of countries introducing registries or bureaus. The average effect of the introduction of an information sharing institution is picked up by the constant term. In the sample, there are only 32 episodes of changes in the creditor rights index (see Appendix B). This result is consistent with the evidence of Tables 2 and 3 of virtually no changes in average creditor rights index by income level or by legal origin over the 25 year period. Unfortunately, many of the 32 changes occur in former communist countries close to the end of the sample, so we cannot compute the average private credit to GDP ratio five or even three years after the reform. We can only compute changes in the dependent variable for 21 observations with five year averages, and for 25 with three year averages. 11 The results, presented in Panel A, show that, for 3 and 5 year windows, a one point increase in the creditor rights score raises the ratio of private credit to GDP by between 6.5 and 8 percentage points. The result (displayed in Figure 4) is statistically significant, and the estimate is quantitatively similar to those in Tables 5 and 6. The control for the average change in the ratio of private credit to GDP among non-reforming countries is not statistically significant. The results also show that, over a 5 year horizon, the private credit to GDP ratio rises after the introduction of both public registries and private bureaus by 7 to 8 percentage points. The magnitudes of the effects are smaller than those reported in Table 6, but the results are statistically significant. However, although parameter estimates show a rise in the private credit to GDP ratio over a 3 year window around the reform, these results are insignificant. Panel B presents the results of an alternative strategy of measuring the effects of reform, namely panel regressions. Here the dependent variable is the ratio of private credit to GDP each year in each sample country, but we introduce both country effects and year effects. Identification here only takes advantage of within country variation in institutional variables. A 11 In a recent paper, Haselmann, Pistor, and Vig (2005) focus specifically on transition economies to examine the effects of changes in creditor protections on bank lending. Their conclusions are broadly consistent with ours. 22

24 key question in this specification is what standard errors to compute. One possibility is to compute robust standard errors. This strategy does not take account of possible serial correlation of error terms within each country. Another possibility is to cluster standard errors by country, as recommended by Bertrand et al. (2004), which imposes no restrictions on the possible serial correlation of error terms. With many time series observations (recall we have 25 years of data), such a specification may blow up standard errors excessively. 12 Since both strategies have problems, we present both standard errors in Table 7B. These panel results yield lower parameter estimates than the findings from the crosssections in Tables 5 and 6 as well as those in Panel A (a 4 percentage point rise in the private credit to GDP ratio per one point improvement in the creditor rights score). The effect is statistically significant for both all and the rich countries using robust standard errors, but only for the rich countries (at the 10 percent level) using standard errors clustered by country. The effects of private bureaus and of public registries remain of the same magnitude as in the earlier specifications, although again standard errors rise sharply when clustered by country. The evidence on the information sharing institutions confirms the previous findings; the evidence on creditor rights is weaker both in magnitude and in statistical significance. Except for the large standard errors clustered by country, the results for reforms, including those from Panel A, support the findings from the cross-section. Both the power and the information channel of credit play a role, and improvements in either channel raise the ratio of private credit to GDP. We recognize that the results in this section are subject to the criticism that the timing of reforms is endogenous, and in particular may occur when policy makers expect private credit to grow. Still, the consistency of results across methodologies is encouraging. 12 In a conversation, James Stock suggests that while clustering of standard errors has a modest impact for 3 to 5 year panels, it often blows up standard errors for longer panels (e.g., 20 years) by a factor of 2 to 3. 23

25 V. Determinants of Credit Institutions We have shown that credit institutions have a significant effect on credit market development. But what shapes these institutions? We have already seen in Tables 2 and 3 that the level of economic development and legal origin are potentially important determinants of which countries have which institutions. In a recent paper, Stulz and Williamson (2003) emphasize that culture is an important determinant of creditor rights. Using the La Porta et al. (1997) sample, they find that creditor rights are weaker in catholic countries, and that holding religious composition constant, legal origin does not help predict creditor rights. Table 8 revisits some of these issues, looking at the potential determinants of creditor rights, public registries, and private bureaus using our larger 2003 sample. With respect to creditor rights, the first column shows that French legal origin countries have sharply weaker rights than the (omitted) common law category without controlling for religion. The second column shows that, consistent with Stulz and Williamson (2003), protestant countries have statistically significantly stronger creditor rights than others, without controlling for legal origin. However, when we put legal origin and religious orientation together in the regression, as we do in columns 3 and 4, religious variables are no longer significant. We also could not get religion to matter when we construct the more complex variables suggested by Stulz and Williamson. The two additional panels in Table 8 examine the determinants of the presence of public registries and private bureaus. French and German legal origin countries are more likely to have public credit registries than their common law counterparts, and transition countries are less likely to have private credit bureaus than the common law countries. Richer countries as well are more likely to have private credit bureaus. Protestant countries are less likely to have public credit registries than others, but the result loses its significance once we control for legal origin. There is no difference between catholic and protestant countries in the incidence of private credit 24

26 bureaus, although muslim and orthodox countries are less likely to have them. In this sample, legal origin appears to matter more for credit institutions than culture. The result that French legal origins predicts the presence of public credit registries, which in turn benefit credit markets in poorer countries, is consistent with the argument of Djankov et al. (2003a) that public ownership and regulation represent solutions to the problem of disorder practiced especially by French civil law countries. Research on the regulation of entry, civil procedure, bank ownership, and the regulation of labor markets has not, however, been able to identify any benefits of this approach to social control of business in developing countries. Perhaps because public registries neither employ significant numbers of people nor can be easily used to discriminate among borrowers for political reasons, they appear to be beneficial. This evidence, then, illustrates a successful public ownership solution to the problem of supporting credit markets in developing countries. VI. Conclusion Using a new sample of credit institutions in 129 countries over 25 years, we revisit some old and establish some new results on private credit around the world. First, in this large sample, both the cross-sectional evidence and the analysis of reforms show that legal creditor rights and information sharing institutions are statistically significant and quantitatively important determinants of private credit development. Second, there is no evidence the creditor rights are converging among legal origins, or among rich and poor countries. This evidence contradicts the legal convergence hypothesis which holds that institutions converge toward the more successful ones over time. Since credit institutions vary so much across countries and legal origins, and since these differences matter for the development of debt markets, this evidence also contradicts the functional convergence hypothesis, which holds that institutions in different countries, while distinct on the surface, functionally converge to accomplish the same goals. 25

27 Third, the data show that the effectiveness of alternative institutions varies systematically across countries at different levels of economic development. Most interestingly, both private and public credit bureaus are strongly associated with private credit in the poorer, but not the richer, countries, suggesting a role for government in facilitating information sharing. 13 Since public credit registries are especially common in French legal origin countries, these results highlight a successful civil law response to financial underdevelopment. Such successes have been difficult to identify in other domains of social control of business. These results suggest that the choice between common and civil law approaches to solving economic problems is a matter of comparative advantage of alternative strategies of social control and not just of the absolute advantage of common law solutions. VII. References Aghion, Philippe, and Patrick Bolton, 1992, An Incomplete Contracts Approach to Corporate Bankruptcy, Review of Economic Studies 59, Barth, James, Gerard Caprio, and Ross Levine, 2004, Bank Regulation and Supervision: What Works Best?, Journal of Financial Intermediation 13, Beck, Thorsten, Ross Levine, and Norman Loayza, 2000, Finance and the Sources of Growth, Journal of Financial Economics 58, Beck, Thorsten, Asli Demirguc-Kunt, and Ross Levine, 2000, A New Database on Financial Development and Structure, World Bank Economic Review 14, Beck, Thorsten, Asli Demirguc-Kunt, and Ross Levine, 2003a, Law and Finance: Why Does Legal Origin Matter?, Journal of Comparative Economics 31, These results echo recent research on the role of free media another mechanism of information sharing in promoting good outcomes in both economic and political markets (Zingales 2000, Djankov et al. 2003b). Unlike in the case of the media, however, our data suggest that information sharing through public firms is beneficial rather than detrimental. A possible reason for the difference is that the scope for public abuse of information collected about the debtors is more limited than that for distortion of news reporting by state-owned media. 26

28 Beck, Thorsten, Asli Demirguc-Kunt, and Ross Levine, 2003b, Law, Endowments, and Finance, Journal of Financial Economics 70, Beck, Thorsten, Asli Demirguc-Kunt, and Ross Levine, 2005, Law and Firms Access to Finance, American Law and Economics Review 7, Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan, 2005, How Much Should We Trust Differences-In-Differences Estimates, Quarterly Journal of Economics 119, Black, Bernard, 2000, Political Determinants of Corporate Governance, Oxford: Oxford University Press. Djankov, Simeon, Edward Glaeser, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, 2003a, The New Comparative Economics, Journal of Comparative Economics 31, Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, 2003b, Courts, Quarterly Journal of Economics 118, Djankov, Simeon, Caralee McLiesh, Tatiana Nenova, and Andrei Shleifer, 2003c, Who Owns the Media, Journal of Law and Economics 46, Glaeser, Edward, Rafael La Porta, Florencio Lopez de Silanes, and Andrei Shleifer, 2004, Do Institutions Cause Growth?, Journal of Economic Growth 9, Hart, Oliver, and John Moore, 1994, A Theory of Debt Based on the Inalienability of Human Capital, Quarterly Journal of Economics 109, Hart, Oliver, and John Moore, 1998, Default and Renegotiation: A Dynamic Model of Debt, Quarterly Journal of Economics 113, Haselmann, Rainer, Katharina Pistor, and Virkant Vig, 2005, How Law Affects Lending, Mimeo, Columbia Law School. 27

29 Jappelli, Tullio, and Marco Pagano, 2000, Information Sharing in Credit Markets: the European Experience, Working Paper 35, University of Salerno, Italy. Jappelli, Tullio, and Marco Pagano, 2002, Information Sharing, Lending, and Defaults: Crosscountry Evidence, Journal of Banking and Finance 26, La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, 1997, Legal Determinants of External Finance, Journal of Finance 52, La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, 1998, Law and Finance, Journal of Political Economy 106, Levine, Ross, 1999, Law, Finance, and Economic Growth, Journal of Financial Intermediation 8, Mulligan, Casey, and Andrei Shleifer, 2005, The Extent of the Market and the Supply of Regulation, Quarterly Journal of Economics 120, Pagano, Marco, and Tullio Jappelli, 1993, Information Sharing in Credit Markets, Journal of Finance 43, Qian, Jun, and Philip Strahan, 2004, How Law and Institutions Shape Financial Contracts: The Case of Bank Loans, Mimeo, Boston College. Sapienza, Paola, 2002, The Effects of Banking Mergers on Loan Contracts, Journal of Finance 57, Stiglitz, Joseph, and Andrew Weiss, 1981, Credit Rationing in Markets with Imperfect Information, American Economic Review 71, Stulz, Rene, and Rohan Williamson, 2003, Culture, Openness, and Finance, Journal of Financial Economics 70, Townsend, Robert, 1979, Optimal Contracts and Competitive Markets with Costly State Verification, Journal of Economic Theory 21, Zingales, Luigi, 2000, In Search of New Foundations, Journal of Finance 55,

30 Table 1: Description of the Variables This table presents the variables used in the regression analysis. The Creditor rights variable updates La Porta and others (1998) and refines the collection methodology. The Public registry and Private bureau variables are newly-collected data. The variable Information sharing is a discrete variable, which aggregates data from the Public Registry and Private bureau variables. The variable Private credit to GDP comes from the IMF's International Financial Statistics. GDP and inflation data come from the World Bank. The data for contract enforcement days is based on the methodology in Djankov and others (2003). The variable Legal origin comes from La Porta and others (1999) and is updated with the CIA Factbook (2003). The variable Religion comes from Stulz and Willaimson (2003) and is updated with the CIA Factbook (2003). All variables are constructed as of January for every year from 1978 to Variable Creditor rights Description An index aggregating creditor rights, following La Porta and others (1998). A score of one is assigned when each of the following rights of secured lenders are defined in laws and regulations: First, there are restrictions, such as creditor consent or minimum dividends, for a debtor to file for reorganization. Second, secured creditors are able to seize their collateral after the reorganization petition is approved, i.e. there is no "automatic stay" or "asset freeze." Third, secured creditors are paid first out of the proceeds of liquidating a bankrupt firm, as opposed to other creditors such as government or workers. Finally, if management does not retain administration of its property pending the resolution of the reorganization. The index ranges from 0 (weak creditor rights) to 4 (strong creditor rights) and is constructed as at January for every year from 1978 to Public registry Private bureau The variable equals 1 if a public credit registry operates in the country, 0 otherwise. A public registry is defined as a database owned by public authorities (usually the Central Bank or Banking Supervisory Authority), that collects information on the standing of borrowers in the financial system and makes it available to financial institutions. The variable is constructed as at January for every year from 1978 to The variable equals 1 if a private credit bureau operates in the country, 0 otherwise. A private bureau is defined as a private commercial firm or non profit organization that maintains a database on the standing of borrowers in the financial system, and its primary role is to facilitate exchange of information amongst banks and financial institutions. Private credit reporting firms, which collect information from public sources but not banks and financial institutions, operate in several other countries but are not considered here. The variable is constructed as at January for every year from 1978 to Information sharing Private Credit to GDP The variable equals 1 if either a public registry or a private bureau operates in the country, 0 otherwise, and is constructed as at January for every year from 1978 to Ratio of credit from deposit taking financial institutions to the private sector (IFS lines 22d and 42d) to GDP (IFS line 99b),expressed as a percentage. Line 22d measures claims on the private sector by commercial banks and other financial institutions that accept transferable deposits such as demand deposits. Line 42d measures claims on the private sector given by other financial institutions that do not accept transferable deposits but that perform financial intermediation by accepting other types of deposits or close substitutes for deposits (e.g. savings and mortgage institutions, post office savings institutions, building and loan associations, certain finance companies, development banks and offshore banking institutions). Source: IMF International Financial Statistics September GDP Logarithm of gross national income (current U.S. Dollars), average Source: World Development Indicators GDP per capita Logarithm of gross national income per capita (Atlas method), Source: World Development Indicators GDP per capita growth Average annual growth in gross domestic product per capita from Source: World Development Indicators Inflation Annual percentage inflation, GDP deflator, average Source: World Development Indicators Contract enforcement days The number of days to resolve a payment dispute through courts. The data are based on the methodology in Djankov and others (2003) but describe the number of calendar days to enforce a contract of unpaid debt worth 50% of the country's GDP per capita. The variable is constructed as at January Legal origin Religion A dummy variable that identifies the legal origin of the Company law or Commercial Code of each country. The five origins are English, French, German, Nordic and Socialist. Source: La Porta and others (1999) and the CIA Factbook A dummy variable that identifies the religion practiced by the largest proportion of the population. There are nine religions: Athiest, Buddhist, Catholic, Hindu, Indigenous, Judaism, Muslim, Orthodox Christian and Protestant. Source: Stulz and Williamson (2003) and the CIA Factbook 2003.

31 Table 2: Credit Institutions by Legal Origin This table presents the means of the Creditor rights, Public registry, Private bureau and Information sharing variables at five year intervals between 1978 and The data are shown by legal origin and T-tests on the differences between English and French legal origins are provided. The analysis covers 129 countries. A. Creditor Rights Legal Origin English French German Nordic Socialist All T-Test, English vs French a a a a a a Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level B. Public Registry Legal Origin English French German Nordic Socialist All T-Test, English vs French a a a a a a Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level C. Private Bureau Legal Origin English French German Nordic Socialist All T-Test, English vs French Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level D. Information Sharing Legal Origin English French German Nordic Socialist All T-Test, English vs French a a a b a Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level

32 Table 3: Credit Institutions by Income Group in 1978 This table presents the means of the Creditor rights, Public registry, Private bureau and Information sharing variables at five year intervals between 1978 and The data are shown by income group at the beginning of the sample period (1978). The sample is split into two, with rich countries having income per capita above the median value and poor countries having income per capita below the median value. T-tests on the differences between rich and poor countries are provided. The analysis covers 129 countries. A. Creditor Rights Income Group Poor Rich Missing 1978 GDP data All T-Test, Rich vs Poor a a a b a a Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level B. Public Registries Income Group Poor Rich Missing 1978 GDP data All T-Test, Rich vs Poor b b b Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level C. Private Bureaus Income Group Poor Rich Missing 1978 GDP data All T-Test, Rich vs Poor a a a a a a Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level D. Information Sharing Income Group Poor Rich Missing 1978 GDP data All T-Test, Rich vs Poor a a a c b Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level

33 Table 4: Correlations This table presents Bonferroni-adjusted simple correlations between all explanatory variables used in the analysis. The definition of each variable is shown in Table 1. The analysis covers 129 countries. Creditor rights Private credit to GDP a (0.001) Creditor rights Public registry Private bureau Information sharing GDP GDP per capita growth Contract enforcement days Public registry c a (0.063) (0.011) Private bureau a c a (0.000) (0.061) (0.001) Information sharing a a a (0.000) (0.224) (0.000) (0.000) GDP a c a a (0.000) (0.086) (0.158) (0.000) (0.001) GDP per capita growth a a a a (0.000) (0.357) (0.347) (0.000) (0.005) (0.002) Contract enforcement days a a a a c (0.000) (0.137) (0.003) (0.010) (0.649) (0.001) (0.057) English legal origin c a a c (0.081) (0.003) (0.000) (0.220) (0.074) (0.456) (0.640) (0.155) French legal origin b a a a c c a a (0.018) (0.000) (0.000) (0.225) (0.002) (0.062) (0.061) (0.012) (0.000) German legal origin b b c a a a a (0.029) (0.032) (0.483) (0.191) (0.096) (0.014) (0.000) (0.803) (0.005) (0.000) Nordic legal origin b b b c a b (0.037) (0.934) (0.030) (0.015) (0.309) (0.058) (0.509) (0.006) (0.219) (0.051) (0.426) Socialist legal origin a a a a b a b a (0.010) (0.247) (0.014) (0.003) (0.000) (0.042) (0.006) (0.802) (0.035) (0.001) (0.173) (0.546) GDP per capita a a c a a a a a c a a b (0.000) (0.006) (0.099) (0.000) (0.000) (0.000) (0.010) (0.000) (0.655) (0.082) (0.000) (0.006) (0.039) Catholic c a a c a a b a (0.241) (0.169) (0.088) (0.001) (0.010) (0.073) (0.383) (0.150) (0.000) (0.000) (0.477) (0.170) (0.019) (0.010) Protestant a b a a a a a a a (0.015) (0.018) (0.000) (0.009) (0.485) (0.157) (0.827) (0.000) (0.000) (0.000) (0.682) (0.000) (0.160) (0.140) (0.001). Muslim b a c b b a a a (0.036) (0.427) (0.141) (0.000) (0.074) (0.037) (0.766) (0.129) (0.585) (0.299) (0.027) (0.219) (0.473) (0.002) (0.000) (0.004). Buddhist a b b b (0.125) (0.769) (0.381) (0.216) (0.404) (0.409) (0.001) (0.041) (0.830) (0.236) (0.115) (0.566) (0.838) (0.974) (0.025) (0.182) (0.046) English legal origin French legal origin German legal origin Nordic legal origin Socialist legal origin Ln GDP per capita Catholic Protestant Muslim Note: p-values are in parentheses; a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level

34 Table 5: Private Credit to GDP Regressions This table presents cross-section OLS regressions, using a sample of 129 countries. The dependent variable is Private Credit to GDP, averaged over the period. Results are shown separately for poor and rich countries. Poor countries are countries with income per capita below the sample's median. Rich countries are countries with income per capita above the sample's median. Four controls are used in the analysis: total GDP, per-capita income growth, a proxy for the efficiency of the legal system, and inflation. The Inflation variable is averaged over the period. Robust standard errors are in parentheses. Independent Variables Dependent Variable: Private Credit to GDP All Countries Poor Countries Rich Countries GDP a a a a a a (1.254) (1.277) (1.170) (1.071) (2.073) (2.185) GDP per capita growth c a b b (1.081) (1.068) (0.263) (0.821) (2.389) (2.420) Contract enforcement days a a b b (3.249) (3.379) (5.573) (5.365) (3.642) (3.858) Inflation c b b a a (0.061) (0.053) (0.025) (0.023) (0.285) (0.333) Creditor rights a a c a a (2.070) (2.472) (1.270) (1.452) (2.963) (3.410) French legal origin c (6.113) (4.710) (8.372) German legal origin c (9.571) (16.977) (10.373) Nordic legal origin (dropped) (9.530) (10.425) Socialist legal origin a (6.648) (4.024) (10.755) Constant a a c a a (36.014) (37.867) (40.025) (38.442) (56.062) (60.603) Obs R-sq Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level

35 Table 6: Private Credit to GDP Regressions, with Information Sharing This table presents cross-section OLS regressions, using a sample of 129 countries. The dependent variable is Private Credit to GDP, averaged over the period. Results are shown separately for poor and rich countries. Poor countries are countries with income per capita below the sample's median. Rich countries are countries with income per capita above the sample's median. Private bureaus and public registries established after 1999 are classified as 0 in the Private bureau and Public registry variables. Information sharing is a discrete variable equal to 1 is either a private bureau or a public registry exists, 0 otherwise. Four controls are used in the analysis: total GDP, per-capita income growth, a proxy for the efficiency of the legal system, and inflation. The Inflation variable is averaged over the period. Robust standard errors are in parentheses. Independent Variables Dependent Variable: Private Credit to GDP All Countries Poor Countries Rich Countries GDP a a a a a a (1.262) (1.403) (1.014) (1.179) (2.222) (2.382) GDP per capita growth b b a a b c (0.999) (0.928) (0.247) (0.224) (2.474) (2.513) Contract enforcement days a a b b (3.170) (2.994) (5.428) (5.287) (3.538) (3.603) Inflation a a (0.070) (0.058) (0.035) (0.032) (0.292) (0.304) Creditor rights a a c c a a (2.116) (2.139) (1.328) (1.272) (2.985) (3.021) Information sharing a a (4.713) (3.284) (7.292) Private bureau a b (6.225) (5.892) (7.131) Public registry a (4.588) (3.616) (6.696) Constant a a a a (35.429) (38.764) (37.678) (34.995) (57.949) (64.304) Obs R-sq Note: a=significant at the 1% level, b=significant at the 5% level, c=significant at the 10% level

36